automated engineering quiz contest system using wi-fi and … · (iecep) use the last type contest...

Presented at the DLSU Research Congress 2016

De La Salle University, Manila, Philippines

March 7-9, 2016

Title of Manuscript, 14 pt. Century font

John Surname1, Esther Surname 2 and Author Three1,* (10 pt. Century) 1 Affiliation (8 pt. Century) 2 Affiliation (8 pt. Century)

*Corresponding Author: email [email protected]

Abstract: A brief summary of approximately 250 – 300 words outlining the

background, objectives of the study, the methodology used, key results and

conclusions. Manuscripts must be submitted as Word files using the prescribed

format. The final paper (extended abstract/full paper) should be no more than six

pages in length. Save the manuscript as MS word document and name the file as:

Research theme CODE of the paper_Surname of Corresponding Author_First name

initial.(dot)Middle initial_PaperX (if corresponding author is submitting more than 1

paper)_VersionY (in cases when you update a former submission)_full paper.

Example: FNH_Santos_J.C_full paper

WCF_Cruz_N.T_Paper2_full paper

HCT_Vidal_M.R_Paper1_version2_full paper

Key Words: up to five key words/terms; separated by semicolons

1. SECTION

1.1 Subsection Main text/Content font: Century. Font size:

Section (12 pt), subsection (11 pt), Sub-subsection (10

pt), Main text (9 pt). For the main text (contents) and

references, two column formatting is required.

In general, papers will have sections for the

introduction, methodology, results and discussions,

acknowledgements, conclusions and

recommendations. However, authors may exercise

some flexibility in organizing the content of their

papers. The paper should be formatted with 1-inch

side margins, 1.5-inch top margin and 1-inch bottom

margin in standard letter size (8.5” x 11”). The whole

paper should be single-spaced and justified. The

extended abstract may contain figures, tables and/or

images and should not exceed 4 pages including

references. The pages should be numbered.

The introduction should give a brief

background of the study, describe relevant

developments in the literature to date, and describe

the objectives and scope of the study. Indent

paragraphs by 0.5-inch.

2. METHODOLOGY

This section should present the methodology

utilized in the study in order to achieve the

objectives. Figures and tables should be referred to in

the text. They should be centered as shown below

and must be of good resolution. Where equations are

used, adequate definition of variables and

parameters must be given, as shown in the example

below.

Wi + j rij = Sii (Eq. 1)

Automated Engineering Quiz Contest SystemUsing Wi-Fi and Android App

Ann E. Dulay∗, Katrina M. Pangan †, Mary L. Requilman‡, Mark U. Te§ and Bryan L. Young¶∗Electronics and Communications Engineering Department

De La Salle UniversityEmail: [email protected]

† Email: katrina [email protected]‡ Email: mary rose [email protected]§ Email: mark kevin [email protected]¶ Email: bryan [email protected]

Abstract—This research aims to develop a Wi-Fi baseddata transmission system for automated engineering quizcontest applications. The system consists of an Androidapp that serves as the ’answer sheet’ for each schoolparticipant. The recognition system used is the OpticalCharacter Recognition (OCR). The system mimics thetypical engineering quiz contest where questions areflashed on the screen and the contestants are givena certain amount of time to give their answer. Thesystem was tested with fifteen devices operated by fifteenparticipants in order to determine the capacity of thenetwork to simultaneously process the transmitted signal.The timer flashed on the screen is synced to the timer ofthe image-to-text converter and a grace period is allottedto allow all the images to get through. The grace periodcan be set by the administrator depending on the numberof participants. For the fifteen devices tested, a 1-secgrace period is added to the timer of the converter sincethe time it takes to transmit the data from the Androiddevice to the server is 0.1 sec. When this time lapses, theconverter program closes thereby blocking all the othertransmissions. The test shows that this time is sufficientto accommodate all fifteen devices.

The OCR has a character recognition accuracy of83.95 %. The accuracy of detection for answers compris-ing several letters and symbols depends entirely on theway the letters were written. If a letter overlaps anotherletter, the OCR misinterprets it to another character,hence the image is converted incorrectly. It is thereforerecommended that scoring be done manually but it’s GUIis still embedded with the GUI of the ’answer sheets’ sothat as each answer is shown, the scorer simply pressesa GUI button to update the score on the correspondingcell.

The communications between the Android device andthe server is simplex, hence, the system has no controlover the Android device. It is recommended that the Appof the Android device be integrated with a timer that willbe synced with the server. This would require two way-communications and allows the system to have controlof the Android device such that when the time lapses,the App is disabled, hence disallowing the device from

sending the answer to the server. This would ultimatelyeliminate disputes regarding the ability of the system toreceive the answers at the allotted time.

I. INTRODUCTION

Quiz contests come in different forms. Some con-tests are based on the speed of the contestants togive the correct answer, that is, whoever presses thebuzzer first shall be the one who will give the answer.This type is practical for contests with less thanfive contestants. Other contests use devices that thecontestants press to give their answers ([1], [2], [3],[4]), but this type is practicable only for contests whereanswers are supplied through multiple choices and themode of answering is by pressing the ’button’ thatcorresponds to one of the choices, either electronicor through a device. Some contests require all of theparticipants to give their answers within the allottedtime, each answer is scored manually and tallied, andwhoever has the highest score at the end of the lastround wins. Engineering quiz contests conducted bythe Institute of Electronics Engineers of the Philippines(IECEP) use the last type contest for its NationalQuiz Bee. It is participated by various schools allover the country with more than thirty contestantsstill remaining after doing the regional elimination.The answers to the questions being asked are notsupplied through multiple choices or other similarforms, instead the contestants write their answers inan acetate because handwritten answers are faster thantypewritten answers ([5]) especially when it involvesanswers from solutions.

During the contest proper, the questions are flashedon the screen; a Quiz master reads the question twice,then a timer is activated. Once the time is up, thecontestants should stop writing, gives the acetate to the

VCR

Typewritten Text

Proceedings of the DLSU Research Congress Vol 4 2016 ISSN 2449-3309



March 7-9, 2016
















1. SECTION






















2. METHODOLOGY








below.


collector, the collector brings the acetate to the OHPoperator, and the answer of each contestant is flashedon the screen, and another person writes and talliesthe scores. This process requires several people doingthe collecting, the OHP operation and the scoring, andconsequently requires lots of time to do all of theseprocesses. This research aims to remove the need forsuch manual process by using an Android device anda Wi-Fi connection. The Android device serves as thecontestants answer sheets through the use of a ScratchPad Application. The answers are sent to the MasterUnit (MU) through Wi-Fi. The MU contains the mainprogram that has several GUIs for showing the imagesof the answers and the running total of scores. Theconversion of the handwritten image into text is imple-mented using an optical character recognition (OCR)algorithm. A dedicated router is set-up to create thelocal area network for the Android tablets. Each tabletis registered before the start of the contest where eachcontestant is provided the password of the router sothat other Android devices in the venue would notmeddle with the entire quiz contest process.

The rest of the paper is constructed as follows.Section II explains the quiz contest setup. In SectionIII, the process flow of the system along with thedeveloped graphical user interface (GUI), the imple-mented optical character recognition (OCR) system,and the developed Scratch Pad Writing Applicationare described. Section IV presents the evaluations anddiscussions based on the results. Finally, Section Vprovides some concluding remarks and offers somerecommendations for future work.

II. QUIZ CONTEST SET-UP

The quiz contest set-up is shown in Fig.1. It com-prises of the Android devices, the router, and the mainserver. For purposes of demonstration, only fifteen (15)Android tablets are utilized in the simulation of thecontest. The limitation is due to the available devicesrather than the capacity of the router. Each Androiddevice is first registered by connecting to the dedicatedlocal router Wi-Fi. Before the registration process,the devices are installed with the Scratch Pad appdeveloped by the proponents. Fig. 2 shows the userinterface of the android app that serves as the answersheet. It contains the pen, the text tool, eraser tool, andthe worksheet shown on top and at the bottom is themail icon which serves as the SEND button. The firstthree tools can be adjusted depending on the androiddevice’s settings. The router is communication mediabetween devices and the main server.

answers sent by the participants. A mainframe served as the master unit andclassrooms, conference halls, contest sites,was conducted to enable users to experiencemore accessible way of data sharing by nology. The rest of the paper is constructed as follows.explains the quiz contest setup. In Sectionflow of the system along with the developeduser interface (GUI), the implementedrecognition (OCR) system, and the developedWriting Application are described. Sectionthe evaluations and discussions based onnally, Section V provides some concludingoffers some recommendations for future II. QUIZ CONTEST SETUP In order to determine the results of thesearchers conducted a quiz contest with(15) participants wherein each participantan Android tablet where the developedtion was installed. The Android applicationthe participants to be connected to the masterFi. Once the connection has been verified,data transmission can already be established.were then flashed in the screen once thestarted. The contestants were then requiredfinal answer on the tablet and press thebefore the designated time ran out. Oncebeen sent, the master unit will then automaticallythe answers through the automated systemautomatically update the score sheet. III. SYSTEM PROCESS FLOW A wireless network using the D-Link DIRGigabit Router is established in order tofifteen (15) simultaneous connections ofAn Android application that serves as thethe connection entry point to the networkThis application is also developed to functionswer sheet or writing canvas of the participants.from the Android application in the formthen transmitted to the master unit throughcation. A database is also developed somaster unit to compile all of the transmittedall of the devices connected to the network. Different MATLAB *.fig files consistquiz contest. The quiz questionnaire, poster,and list of the contestants are created quiz master provided that the three files

central computer and was installed in sites, etc. This study

experience an easier and using Wi-Fi tech-

follows. Section II Section III, the process

developed graphical implemented optical character

developed Scratch Pad Section IV presents

on the results. Fi-concluding remarks and

work.

the study, the re-with at least fifteen

participant was provided developed Android applica-

application then enabled master unit via Wi-

verified, the one-way established. Questions

the quiz contest has required to write the the “submit” button Once the data have automatically verify

system developed, and

DIR-655 Wireless N to connect at least

of Android devices. the verification and

network is developed. function as the an-

participants. The data form of *.jpg files are

through a server appli-so as to allow the

transmitted data and view network.

consist the GUI for the poster, mechanics, and edited by the are saved as *.jpg

files per slide following the(i.e. Slide1.jpg Fig. 1. The process flow of the system. Fig. 2. The user nterface of the developedtion. will contain question numbernected to the Score Databasewhere the scores gatheredand scores displayed to eachand retrieved, respectively.the number of participantsassigned duration of the timerwhich ranges from 30 seconds,conds. The GUI then callsble to running the *.m filesscoring of images at the endof each question, the answerswith their registered names,scores are displayed. The option to manually overridequestion. The ranking ofplayed at the end of the quiz

the file allocation for the GUI

system.

developed ScratchPad writing applica-

number one). The latter is then con-Database which is the same file

gathered from each question windows each answer windows are stored

respectively. The scaling of the GUI limits participants to only 15-30. The GUI has an

timer depending on the category seconds, 60 seconds, and 90 se-

calls the function which is responsi-files for automatic checking and end of each question. At the end

answers of the contestants along names, current scores, and total

developed GUI also allows the override the scored at the end of every

of the contestants are then dis-quiz contest.

Fig. 1. Quiz Contest Set-up

The summary of the contest process with the auto-mated system in place is shown in Fig. 3. It startswith the installation of Android application to thetablets of each contestant. Each participant is thenasked to establish the connection with the dedicatedrouter by providing them a password. This is done toprevent other devices of the audience from connectingwith the dedicated Wi-Fi. As each participant connectstheir devices, the Administrator/Quiz Master writesthe name of the school in the GUI. Once all deviceshave been registered, the contest proper starts. Eachquestion is flashed on the screen; timer starts after thequiz master reads the question twice. The contestantsshould write their final answer on the tablet and pressthe submit button before the designated time ran out.Once the data have been sent, the master unit willthen automatically verify the answers through theautomated system developed which converts the image(handwritten text) to text, and automatically update thescore sheet. The original handwritten answer of eachcontestant is flashed on the screen while the scores areautomatically updated and shown at a portion on theside of the screen.

III. SYSTEM PROCESS FLOW

A wireless network using the D-Link DIR-655Wireless N Gigabit Router is established in order toconnect at least fifteen (15) simultaneous connectionsof Android devices. An Android application that servesas the verification and the connection entry point tothe network is developed. This application is alsodeveloped to function as the answer sheet or writingcanvas of the participants. The data from the Androidapplication in the form of *.jpg files are then transmit-ted to the master unit through a server application. Adatabase is also developed so as to allow the master

VCR

Typewritten Text

VCR

Typewritten Text

VCR

Typewritten Text




March 7-9, 2016
















1. SECTION






















2. METHODOLOGY








below.


answers sent by the participants. A mainframe served as the master unit andclassrooms, conference halls, contest sites,was conducted to enable users to experiencemore accessible way of data sharing by nology. The rest of the paper is constructed as follows.explains the quiz contest setup. In Sectionflow of the system along with the developeduser interface (GUI), the implementedrecognition (OCR) system, and the developedWriting Application are described. Sectionthe evaluations and discussions based onnally, Section V provides some concludingoffers some recommendations for future II. QUIZ CONTEST SETUP In order to determine the results of thesearchers conducted a quiz contest with(15) participants wherein each participantan Android tablet where the developedtion was installed. The Android applicationthe participants to be connected to the masterFi. Once the connection has been verified,data transmission can already be established.were then flashed in the screen once thestarted. The contestants were then requiredfinal answer on the tablet and press thebefore the designated time ran out. Oncebeen sent, the master unit will then automaticallythe answers through the automated systemautomatically update the score sheet. III. SYSTEM PROCESS FLOW A wireless network using the D-Link DIRGigabit Router is established in order tofifteen (15) simultaneous connections ofAn Android application that serves as thethe connection entry point to the networkThis application is also developed to functionswer sheet or writing canvas of the participants.from the Android application in the formthen transmitted to the master unit throughcation. A database is also developed somaster unit to compile all of the transmittedall of the devices connected to the network. Different MATLAB *.fig files consistquiz contest. The quiz questionnaire, poster,and list of the contestants are created quiz master provided that the three files

central computer and was installed in sites, etc. This study

experience an easier and using Wi-Fi tech-

follows. Section II Section III, the process

developed graphical implemented optical character

developed Scratch Pad Section IV presents

on the results. Fi-concluding remarks and

work.

the study, the re-with at least fifteen

participant was provided developed Android applica-

application then enabled master unit via Wi-

verified, the one-way established. Questions

the quiz contest has required to write the the “submit” button Once the data have automatically verify

system developed, and

DIR-655 Wireless N to connect at least

of Android devices. the verification and

network is developed. function as the an-

participants. The data form of *.jpg files are

through a server appli-so as to allow the

transmitted data and view network.

consist the GUI for the poster, mechanics, and edited by the are saved as *.jpg

files per slide following the(i.e. Slide1.jpg Fig. 1. The process flow of the system. Fig. 2. The user nterface of the developedtion. will contain question numbernected to the Score Databasewhere the scores gatheredand scores displayed to eachand retrieved, respectively.the number of participantsassigned duration of the timerwhich ranges from 30 seconds,conds. The GUI then callsble to running the *.m filesscoring of images at the endof each question, the answerswith their registered names,scores are displayed. The option to manually overridequestion. The ranking ofplayed at the end of the quiz

the file allocation for the GUI

system.

developed ScratchPad writing applica-

number one). The latter is then con-Database which is the same file

gathered from each question windows each answer windows are stored

respectively. The scaling of the GUI limits participants to only 15-30. The GUI has an

timer depending on the category seconds, 60 seconds, and 90 se-

calls the function which is responsi-files for automatic checking and end of each question. At the end

answers of the contestants along names, current scores, and total

developed GUI also allows the override the scored at the end of every

of the contestants are then dis-quiz contest.

Fig. 2. Scratch Pad User Interface

Installation of

Android application

to Tablets

Connection of

Tablets to the

Network

Registration of

participants names

in the GUI

Start of Contest

Proper

Flashing of

Questions on the

Screen

Answering of the

questions using

Android App scratch

pad

Sending of answers

to Master Unit via

Wifi

Conversion of

handwritten image

to text

Automatic Scoring

Displaying of

Answers and scores

on the screenA

A

Fig. 3. The Contest Process

unit to compile all of the transmitted data and view allof the devices connected to the network.

Different MATLAB *.fig files consist the GUI forthe quiz contest. The quiz questionnaire, poster, me-chanics, and list of the contestants are created andedited by the quiz master provided that the three filesare saved as *.jpg files per slide following the fileallocation for the GUI (i.e. Slide1.jpg will containquestion number one). The latter is then connectedto the Score Database which is the same file where

the scores gathered from each question windows andscores displayed to each answer windows are storedand retrieved, respectively. The scaling of the GUI lim-its the number of participants to only 15-30. The GUIhas an assigned duration of the timer depending on thecategory which ranges from 30 seconds, 60 seconds,and 90 seconds. The GUI then calls the function whichis responsible to running the *.m files for automaticchecking and scoring of images at the end of eachquestion. At the end of each question, the answersof the contestants along with their registered names,current scores, and total scores are displayed (Fig. ??)).The developed GUI also allows the option to manuallyoverride the scores at the end of every question. Theranking of the contestants are then displayed at theend of the quiz contest. Fig. 4 summarizes the systemprocess flow.

An OCR program is also implemented in order toconvert the handwriting of the participants to text forthe automatic scoring. In order to do so, the programloads the images sent by the contestants to MATLAB.Once loaded, the image is converted to gray scale, thento black and white. An example is shown in Fig. 6.Several template files containing all of the characters(in matrix form) were used to train the system. Afteridentification of the number of lines and the number ofcharacters in each line, the characters are then croppedindividually and resized to a 42 x 24 matrix.

The correlation between the templates and the inputimage is done by a loaded program. The output of thisprogram is the string containing the letter that the inputimage correlated with. After conversion, the automaticscoring program is then implemented which comparesthe file containing the converted handwriting to thetext file containing the correct answer. The scores arethen stored in an excel file and called by the GUI fordisplay.

IV. RESULTS AND DISCUSSION

A. Device Connectivity

Fig. 7 shows the signal strength of the dLink routeras a function of distance. The tool used to measurethis is the Wifi Analyzer application which is loaded inthe device under test (DUT). The DUT is an Androiddevice and represents all the possible Android devicesthat could be used in the contest. Expectedly, the signalstrength monitored by the DUT gets weaker as it goesfarther from the dLink router, or simply put, the signalstrength is inversely proportional to distance.

This test also shows the reliability of the routerconnection amidst several wifi sources present in the

VCR

Typewritten Text




March 7-9, 2016
















1. SECTION






















2. METHODOLOGY








below.


4.2.1.1 General Flowchart

The GUI is composed of ten GUIDE figure files. Five of these files are

composed Tab Panels that separates the questions and answers tabs for the

Sample Data, Easy, Average, and Difficult

Home window displays the quiz contest details. The Admin edit window is

4.2.1.1 General Flowchart

4.2 General Flowchart of the GUI



Sample Data, Easy, Average, and Difficult levels, and Clincher round. The


82



levels, and Clincher round. The


Fig. 4. System Process Flow

Fig. 5. GUI for Displaying the Answers and Scores

Fig. 6. Character recognition process

area that are also received by the DUT. This is crucialduring the actual contest such that Android devicesregistered by the participants should never be discon-nected at any time during the entire contest.

Figure 5.1 Signal strength vs Distance graph

Figure 5.1 shows the graph of the data obtained from the dLink

It can be seen that as distance increases, the signal strength decreases

suggesting an inverse relationship. The signal strength of other access points

and routers were measured. At distances 8m to 15m, the signal strength of

myWiFi is stronger th

closer to the router of myWiFi at this point. However, even if myWiFi has a

stronger signal strength, that did not become a cause for the Android phone to

be disconnected from the dLink router. There w

the phone was able to scan more than one available network but did not

Figure 5.1 Signal strength vs Distance graph

Figure 5.1 shows the graph of the data obtained from the dLink




myWiFi is stronger than that of the dLink router because the researcher was



be disconnected from the dLink router. There were also several points where


102

Figure 5.1 shows the graph of the data obtained from the dLink router.




an that of the dLink router because the researcher was



ere also several points where


Fig. 7. Wi-fi Signal Strength

cover. From the figures beyond 40m, it can be seen that the other access points

are becoming stronger than the signal of dLink but still did not become

for disconnection. The Android phone was only disconnected from the router

at 43m when the router was already out of reach.

5.1.2 Data Transmission

5.1.2.1 Data Transmission Speed at Varying Distances with Single Device

The set up used for this test is the same as the previous section. This

time, instead of just measuring the signal strength, a file is sent of the server

via the dLink router. This test aims to determine the transmission time it take

for the same file size

Android phone under test and the server.

Figure 5.2 Transmission Time vs Distance Graph

Figure 5.2 shows that no matter what the signal strength, the data being

sent is received by the router

connectivity which is 40 m. This data transmission is made possible because of


are becoming stronger than the signal of dLink but still did not become


at 43m when the router was already out of reach.

5.1.2 Data Transmission

5.1.2.1 Data Transmission Speed at Varying Distances with Single Device




for the same file size to be received at the server at various distances of the

Android phone under test and the server.



sent is received by the router until the device reaches the optimal distance for


105


are becoming stronger than the signal of dLink but still did not become a cause


5.1.2.1 Data Transmission Speed at Varying Distances with




to be received at the server at various distances of the



until the device reaches the optimal distance for


Fig. 8. Transmission Time vs Distance

B. Data Transmission

In this test, the DUT sends data to the router atvarious distances to determine the maximum distanceat which the router could still establish a connection.In the test conducted, the maximum distance at whichthe router can receive data from the Android deviceis 40 m. For the same set-up, the transmission time ismeasured. This is done to determine the grace periodthat will be included in the GUI’s timer before it closesand prohibits any data from any device to be sent tothe main program. Engineering Quiz contest are buzzerbeater types of contests where contestants are requiredto submit their answers within the designated time. Itwas observed in the test that the average time it takes totransmit the data is 0.1 sec and this is notwithstandingthe distance of the DUT from the router as long as itis within its range. The transmission time is relativeto the image size (answers from the Android device).The typical size of the image is only about 30 kb insize.

C. Character Accuracy of OCR

The group aimed to determine the accuracy ofthe OCR system by getting thirty samples per letter.Figs. 9 and 10 show the accuracy obtained per letter,and the letters that the developed system gets confusedwith. Fig. 9 shows the accuracy obtained by each of thecapital letters. The letters which got a perfect accuracywere letters L, and V. Letters which got 90%-99%

VCR

Typewritten Text




March 7-9, 2016
















1. SECTION






















2. METHODOLOGY








below.


The group aimed to determine the accuracy

system by getting thirty samples per letter.show the accuracy obtained per letter, andthe developed system gets confused withFig. 5. The transmission time vs. distance graph.

Fig. 6. The percentage accuracy of capita letters. Fig. 7. The percentage accuracy of small letters.

Fig. 8. The percentage accuracy of numbers. Fig. 9. The percentage accuracy of mathematical symbols.

accuracy of the OCR

letter. Figures 6 – 9 and the letters that

with

symbols.

Table 2 shows the averagesmall letters, as well as theall, the overall accuracy There are several reasonssion from scribble to text sons is the difference betweensample and the availablewere trained. Figure 6 shows the accuracytal letters. The letters whichletters L, and V. Letters whichwere C, D, M, O, S, T, W,- 89% accuracy were A, which got 70% - 79% accuracyZ. Letter R had the lowest was followed by letter I withracy of the capital letters is Figure 7 shows the accuracyter which obtained a perfectwhich obtained 90% - 99%and x. Letters which obtaineda, b, f, h, m, p, and q . Letters

TABLEOVERALL INDIVIDUAL

accuracy were g, t, u, y, andleast accuracy with 50% wereletters e, and k with only of the small letters is 80.38%. Figure 8 shows the accuracyThe number which got 100%Numbers which obtained 90%3. Numbers which got 80 -9. The numbers which got

average accuracy of the capital and the numbers and symbols. All in of the OCR system is 82.4%.

reasons why mismatch in the conver- occurred. One of the main rea-

between the handwriting of the available templates in the system that

accuracy obtained by each of the capi-which got a perfect accuracy were

which got 90% - 99% accuracy W, and X. Letters which got 80%

B, G, H, N, P, and Q. Letters accuracy were E, F, J, K, U, Y, and

accuracy with only 60% which with 63.33%. The average accu-is 83.72%.

accuracy of the small letters. The let-perfect accuracy was letter v. Letters

99% accuracy werec, d, j, o, r, s, w, obtained 80% - 89% accuracy were Letters which obtained 70% - 79%

TABLE 2

INDIVIDUAL CHARACTER ACCURACY and z . Letters which obtained the were letters i, and l followed by 66.67%. The average accuracy

80.38%.

accuracy obtained by numbers 0-9. 100% accuracy was number 0. 90% -99% accuracy were 1 and -89% accuracy were 2, 4, 7, and

got 70% -79% accuracy were 5,

Fig. 9. OCR Accuracy for Capital Letters

accuracy were C, D, M, O, S, T, W, and X. Letterswhich got 80%-89% accuracy were A, B, G, H, N, P,and Q. Letters which got 70%-79% accuracy were E,F, J, K, U, Y, and Z. Letter R had the lowest accuracywith only 60% which was followed by letter I with63.33%. The average accuracy of the capital letters is83.72%.

Fig. 10 shows the accuracy of the small letters. Theletter which obtained a perfect accuracy was letter v.Letters which obtained 90%-99% accuracy were c, d,j, o, r, s, w, and x. Letters which obtained 80%-89%accuracy were a, b, f, h, m, p, and q . Letters whichobtained 70%-79% accuracy g, t, u, y, and z. Letterswhich obtained the smallest accuracy at 50% wereletters i, and l followed by letters e and k with 66.67%.

Fig. 11 shows the accuracy obtained for numbers 0-9. The number which got 100% accuracy was number0. Numbers which obtained 90%-99% accuracy were 1and 3. Numbers which got 80-89% accuracy were 2, 4,7, and 9. The numbers which got 70% -79% accuracywere 5, and 6. Lastly, the number which was the leastaccurate with only 67% was number 8.

Fig. 12 shows the accuracy obtained by the mostcommonly used math symbols. The division symbol,/ and the greater than sign, > had 100% accuracy. Thesymbols with 90% accuracy were the plus symbol (+),lambda (λ), and ohm (Ω). The multiplication symbol(∗) had 70% accuracy. Lastly, the symbols with thelowest accuracy 60% were the caret ( ) and the squareroot (√) symbols. The average accuracy obtained was82.50%.

V. CONCLUSION

A new approach in conducting quiz contests wasintroduced in this study. A Wi-Fi based data transmis-sion system was developed which is specifically forquiz contest applications. The whole system consistsof 5 parts namely: the Android application Scratchpadalong with its dedicated Android server, the optical





accuracy of the OCR


with

symbols.













TABLE 2


80.38%.



Fig. 10. OCR Accuracy for Lowercase Letters





accuracy of the OCR


with

symbols.













TABLE 2


80.38%.



Fig. 11. OCR Accuracy for Numbers





accuracy of the OCR


with

symbols.













TABLE 2


80.38%.



Fig. 12. OCR Accuracy for Symbols

character recognition (OCR) system coded in MAT-LAB, the checking and scoring part of the system, thegraphical user interface (GUI) for the whole system,and the Wi-Fi connection serving as the backbonenetwork connecting all of the parts and combiningthem into one whole system with the use of a router.Based on the test results gathered, the system proveditself to be efficient in three aspects; data transmission,character recognition, and automatic scoring.

On average, it takes less than a second for a personto raise his/her answer in the typical quiz contestset-up. From the various tests conducted, the systemproved that the transmission time between the tablet

VCR

Typewritten Text




March 7-9, 2016
















1. SECTION






















2. METHODOLOGY








below.


and the server is less than a second, as well, regardlessof the distance from the router. The character recogni-tion program (OCR) was trained with various strokesand different handwriting to reach a best case accuracyof 82.40%. The characters tested include both theuppercase and lowercase letters, numbers, and specialmathematical symbols. Best case scenario here meansthat the system is able to convert the handwrittenimage to the correct text when the characters donot overlap, and are in block letters and not scripts.It is therefore recommended that an intensive OCRalgorithm be used or developed to consider scriptsand overlaps in the handwritten to text conversion. Asfor the scoring, the system automatically scored andupdated the running total after each conversion. How-ever, the conversion of answers that contain severalcharacters was not very efficient since the OCR hasvery limited training samples. An additional feature,the manual override, is added to the system to allowthe Quiz master to change the score in cases wherethe system has incorrectly converted the answer ora dispute regarding the answer arises and the judgesconsider the disputed answer as correct.

The scope of this research is delimited to the imple-mentation of the system from the Android App to theOCR conversion up to the scoring only. The possibilityof cheating through sending of text messages to theAndroid device or other manner of cheating involvingthe Android device is not included in this research.It is therefore recommended that the Android app bemade more impervious by either blocking it or includea feature that will close the App once another deviceattempts to send a message to it. It is also recom-mended to improve the OCR algorithm in order to fullyrealize the essence of the system for engineering quizcontest. The very reason why handwriting is preferredover typing character streams is to allow writing ofanswers that will involve equations. While these Appsmay already be available, the speed of the contestantin writing the answers would beat the speed of typingor using a text and equation tool.

REFERENCES

[1] STI, “Think quest,” http://www.sti.edu/25th anniversary/pdf/tagisan/ thinkquest mechanics.pdf .

[2] N. Nanjiani, M. Kuhlke, M. Mishra, and B. Ramakrishnananda,“Automated quiz generation system,” May 12 2015, uS Patent9,028,260. [Online]. Available: http://www.google.com/patents/US9028260

[3] Z. Sarker and M. Rahman, “Online intelligent quiz system,”Master’s thesis, BRAC University, 2007.

[4] B. S. Batok, “Multi-platform tv quiz show,” US PatentUS20 150 145 214 A1, 2015.

[5] D. Li, D. Zhang, and Y. Zhang, “Handwritten input for asianlanguages,” US Patent US20 050 276 480 A1, 2005.

VCR

Typewritten Text


automated engineering quiz contest system using wi-fi and … · (iecep) use the last type contest...

Documents